SMSF Financial Australia
A custom ERP for an Australian self managed super fund practice. Every job in the business is a regulated process, so the system was built from the regulation outward rather than from the software inward.
The brief
A specialist SMSF administration practice running on a legacy CRM, a shared drive and a lot of spreadsheets. The knowledge of how a fund actually gets established lived in people’s heads and in a document nobody had opened in a year.
The brief was one system that runs the whole practice. First enquiry to established fund. Established fund to annual return. Annual return to the next one. Plus the investment side, where the same practice runs capital raises for its clients and has to document every one of them to a standard that survives an audit.
The constraint that shaped everything: in financial services the workflow is the compliance. Each step exists because a rule requires it, not because somebody preferred it that way. A system that makes those steps optional is worse than a spreadsheet, because it looks like control while quietly removing it.
The build
The object model came before a single setting was configured.
A person is a contact. A fund is its own record, not a set of fields on a person. A trustee company is a company. A bare trust is a bare trust. An investment product is the offer and an investment holding is what one specific fund actually owns of it.
That separation sounds academic until you hit the first real case. One person is a member of two funds, a director of three companies and a beneficiary of a family trust. One fund has four members, two trustee companies and nine holdings across four products. A flat contact record cannot carry any of that. The moment you force it to, your reporting starts lying.
So the records were modelled as what they are and joined by named associations. A member belongs to a fund. A fund holds investments. A holding joins a fund to a product. Open any record and you can walk the whole structure in either direction.
Pipelines became the operating system.
Lead through to decision. Establishment. Ongoing management. Annual returns. Wind up. Each one a real job moving through real stages, with the stage change as the event everything else listens for.
The establishment process was encoded, not documented.
Setting up a self managed super fund is roughly seventy discrete tasks across eight stages. Apply for the corporate trustee. Lodge the ASIC forms. Open the bank account. Prepare the investment strategy. Flag every signature point in the pack before it goes in the post. Save the Australia Post invoice for the bookkeeper.
Every one of those is now created automatically when a file reaches the stage it belongs to, assigned to the right person, due on the right day, with weekends skipped. Nobody maintains a checklist. Nobody remembers task forty one.
The two gates are the part I am most pleased with.
There are two points where a file must not proceed without a qualified sign off. Approval to proceed with establishment. Approval of the drafted ASIC entity.
The naive version is a task with the word GATE in the title and a hope that nobody drags the card. The built version inverts it: completing the approval task is what moves the file to the next stage. The approval is not a record of the decision, it is the mechanism of the decision. If someone advances a file without it, an alert fires to two people, a review task is raised and it escalates again the next day if it is still unapproved.
That is the difference between a process you can describe to an auditor and one you can show them.
Document generation was rebuilt as blocks.
The investment application form is a long regulated document that changes in three different rhythms. Some of it never changes. Some changes only when the issuer’s details change. Some changes with every new offer. The rest is merged per investor.
Nineteen near identical templates is how that ends up when nobody stops it. Instead the document was broken into static blocks, per issuer blocks, per offer blocks and per investor blocks, assembled at generation time. A new series is a handful of new blocks, not a new document.
Where the platform could not do it, something else did.
A separate service handles filling the source PDFs, because a lodged form is a regulated artefact and has to look exactly as lodged. An eSignature provider was evaluated against the three things Australian law actually requires: identity, reliability and a separate method of verification. It was selected on API fit and cost per envelope rather than brand recognition. An investment calculator and an investment desk replaced the spreadsheets that used to run a capital raise.
The honest rule throughout: use the platform for what it is genuinely good at and bridge out cleanly where it is not. Forcing a CRM to be a document engine produces something nobody can maintain.
The details
Migration. A legacy CRM carrying years of fund, member, company, trust and holding data, mapped object by object rather than lifted field by field. Every record count reconciled against source before anything was trusted.
Permissions built from the regulation. The platform exposes 125 individual permissions. The practice needed six roles, restricted visibility on tax file numbers and the most sensitive financial data separated into its own environment entirely, because field level restriction does not exist on every object. That constraint was found by reading the platform, not by hoping.
Compliance guardrails, checked before anything shipped. Australian sender registration before a single SMS automation goes live, never assuming US defaults. Consent and a working opt out on every messaging path. Deliverability protected with a dedicated sending domain rather than traded for speed.
The real world, modelled. Business hours in the practice’s own timezone, not the server’s. Due dates that skip weekends. Notification times set per person, because a daily task summary that arrives at three in the morning is not a task summary.
Nothing went live untested. Every automation proven on test records before a real client touched it, with the unhappy path proven as well as the happy one. What happens on no reply, on the wrong data, on a client who replies STOP. Silence is a design decision, not an accident.
Closing
A CRM on the surface. A compliance system underneath.